+ . . PROJECT DETAIL . . +
Natural Language to SQL Analytics System
[ ACTIVE ]LLM-powered analytics interface that converts natural language queries into executable SQL.
Developed a natural language analytics system that enables users to query structured datasets using conversational English. The system translates natural language questions into executable SQL queries and returns analytics results directly from relational databases.
Key Contributions
- Built an LLM-powered text-to-SQL pipeline that converts natural language queries into structured SQL statements.
- Implemented schema-aware retrieval using Qdrant vector search to identify relevant tables and columns before SQL generation.
- Developed schema-aware prompting and context injection to improve query accuracy across multi-table relational databases.
- Built a query validation and correction layer to detect malformed SQL and prevent unsafe queries before execution.
- Integrated the system with database execution workflows to run generated queries and return structured analytics results.
- Designed the backend service using Python and FastAPI, enabling scalable API-based interaction with the analytics engine.
Tech Stack
Python, FastAPI, LLMs, Qdrant, SQL, PostgreSQL, NLP, Prompt Engineering, Vector Search
Links
Private – Developed for employer